A General Framework for Fair Regression

TL;DR

This paper introduces a unified framework to incorporate group fairness constraints into various regression models, including kernel methods and decision trees, ensuring fair predictions without increasing computational complexity.

Contribution

It presents a novel, general approach for embedding fairness constraints into multiple regression techniques, applicable to trained models and requiring only training data group labels.

Findings

Fairness constraints can be integrated without increasing model complexity.

The approach applies to models like Gaussian processes, SVMs, neural networks, and decision trees.

Perturbations to models are tightly bounded by the number of leaves in trees.

Abstract

Fairness, through its many forms and definitions, has become an important issue facing the machine learning community. In this work, we consider how to incorporate group fairness constraints in kernel regression methods, applicable to Gaussian processes, support vector machines, neural network regression and decision tree regression. Further, we focus on examining the effect of incorporating these constraints in decision tree regression, with direct applications to random forests and boosted trees amongst other widespread popular inference techniques. We show that the order of complexity of memory and computation is preserved for such models and tightly bound the expected perturbations to the model in terms of the number of leaves of the trees. Importantly, the approach works on trained models and hence can be easily applied to models in current use and group labels are only required on training data.